Error Conditions

The ITPACKV modules assign to the error flag IER certain values to indicate error conditions (or lack of them) upon exiting. Nonzero integer values of the error flag indicate that an error condition was detected. The values it can be assigned and the corresponding meanings are indicated in the following table.

Error Flag		Meaning
${\bf IER} =$	0,	Normal convergence was obtained.
=	1+M,	Invalid order of the system, N.
=	2+M,	Workspace array WKSP(*) is not large enough. IPARM(8)
		is set to the amount of required workspace, NW.
=	3+M,	Failure to converge in IPARM(1) iterations. RPARM(1)
		is reset to the last stopping value computed.
=	4+M,	Invalid order of the black subsystem, NB.
=	201,	Red-black indexing is not possible.
=	401,	There is a zero diagonal element.
=	402,	No diagonal element in a row.
=	501,	Failure to converge in ITMAX function evaluations.
=	502,	Function does not change sign at the endpoints.
=	601,	Successive iterates are not monotone increasing.
=	602,	The matrix is not positive definite.

JCG(), JSI(), SOR(), SSORCG(), SSORSI(), RSCG(), RSSI() assign values to M of 10, 20, 30, 40, 50, 60, 70, respectively. PRBNDX(), SCAL(), ZBRENT(), EQRT1S() are subroutines with error flags in the 200's, 400's, 500's, 600's, respectively. These routines perform the following functions: PRBNDX() determines the red-black indexing, SCAL() scales the system, ZBRENT() is a modified IMSL routine for computing a zero of a function which changes sign in a given interval, EQRT1S() is a modified IMSL routine for computing the largest eigenvalue of a symmetric tridiagonal matrix.⁷